Browse 20 offerings: 14 data lakes plus MCP integrations, a marketplace protocol, and bespoke data services. Record counts below are snapshot figures reported by the Data Foundry catalog; live availability, freshness and licensing are confirmed in writing at quote time.
3.71M
Sum of per-dataset snapshot counts (pending backend verification)
14 Lakes + 6
14 data lakes; 2 MCP integrations, 1 protocol, 3 services
Parquet / JSONL
Native DuckDB, PyArrow, Pandas & Polars
Per dataset
Refresh cadence is stated on each dataset page and confirmed at quote
DATA FOUNDRY CATALOG
Inspect schemas, record counts, and update cadences. Every lake includes sample schema definitions, field dictionaries, and clear commercial scoping pathways.
Over 3.61 million SEC EDGAR corporate filing index rows joined with 10,409 verified public corporate tickers and CIK identifiers. Direct from official SEC EDGAR dissemination feeds.
45,300 drug label and adverse reaction records normalized from the US FDA FAERS repository. Standardized MedDRA reaction terms, dosage data, and patient demographic signals.
19,000 verified academic preprints across artificial intelligence, machine learning, and computation. Sourced via official arXiv OAI-PMH dissemination with full author and citation graphs.
24,933 verified records: multi-chain daily circulating supply time-series (DefiLlama) and market valuation snapshots (CoinGecko) across USDT, USDC, USDe, and major pegged assets. Reported supply & price observations; not an independent reserve audit.
49,246 verified records: 46,719 daily OHLCV spot klines (Binance.US) and 2,527 market valuation snapshots (CoinGecko). Historical price candles only; does NOT contain L2/L3 order-book data. Restricted feeds remain quarantined.
2,581 settled 5-minute Locational Marginal Pricing (LMP) records across CAISO Day-Ahead and Real-Time nodes, joined with weather telemetry and renewable generation mix.
2,369 enriched vulnerability records joining CISA Known Exploited Vulnerabilities (KEV) with NIST NVD CVSS vectors and FIRST EPSS exploit probability scores.
1,783 structured evaluation rows across 460+ open-weights and frontier LLMs. Measures MMLU, GSM8K, HumanEval, TTFT, and sustained generation tokens/sec across verified hardware rigs.
1,600 patent records and NASA NTRS technical reports across quantum computing, semiconductor lithography, and autonomous robotics. Includes CPC hierarchy and assignee graphs.
1,200 synchronized probabilistic contracts across Polymarket and Kalshi. Normalizes implied probability, open interest, 24h volume, and resolution criteria for event trading agents.
674 prime award transactions from USASpending.gov and SAM.gov. Covers DoD, NASA, and DoE contracts with NAICS codes, obligating agency, vendor DUNS/UEI, and award amounts.
450 multi-year macroeconomic time series covering GDP, consumer price inflation, sovereign debt, and FX reserves across 40 OECD and emerging economies from official Open Data APIs.
Normalized scientific papers sample from the OpenAlex scholarly graph. Enriched with MeSH terms, institutional affiliation IDs, DOI identifiers, and open-access licensing flags.
180 high-magnitude seismic events from the USGS Earthquake Hazards Program. Includes hypocenter coordinates, modified Mercalli intensity, and estimated economic exposure figures.
Model Context Protocol tool server adapter for autonomous algorithmic agents. Exposes streaming on-chain token reads and market snapshots over standardized JSON-RPC 2.0 tool calls.
Deterministic Model Context Protocol tool interface for coding assistants. Analyzes dependency vulnerability vectors, EPSS exploit likelihoods, and static workflow definitions.
Cryptographic quote and delivery protocol for autonomous software agents. Emits verifiable SHA-256 delivery receipts and enables deterministic dataset procurement.
Copyright-cleared domain repositories for fine-tuning LLMs and quant neural networks. Fully verified chain-of-title, provenance certificates, and per-token hygiene filters.
Bespoke multi-stage LLM and deterministic programmatic validation pipelines. Verifies synthetic data distributions, cleans corrupted schemas, and eliminates training leakage.
Authorized legal acquisition and deterministic normalization of specialized vertical datasets. We handle primary ingestion, sanitization, schema mapping, and delivery manifests.
HIGH-PERFORMANCE ARCHITECTURE
All Data Foundry lakes are structured into columnar Apache Parquet files with Zstandard compression and embedded dictionary encoding. Ingest directly into DuckDB, Polars, or PyArrow without running expensive database servers.
-- Illustrative query (output values below are example formatting, not a performance claim) SELECT form, COUNT(*) AS total_filings, MIN(filing_date) AS earliest, MAX(filing_date) AS latest FROM read_parquet('sec_filings_index_*.parquet') WHERE form IN ('10-K', '10-Q', '8-K') GROUP BY form ORDER BY total_filings DESC; ┌─────────┬───────────────┬────────────┬────────────┐ │ form │ total_filings │ earliest │ latest │ ├─────────┼───────────────┼────────────┼────────────┤ │ 8-K │ 1,842,109 │ 2018-01-02 │ 2026-11-08 │ │ 10-Q │ 941,204 │ 2018-01-03 │ 2026-11-06 │ │ 10-K │ 836,053 │ 2018-01-04 │ 2026-11-07 │ └─────────┴───────────────┴────────────┴────────────┘ 3 rows in set (0.042 sec) · Peak RAM: 18.4 MB
Every Data Foundry release includes an immutable SHA-256 partition manifest. Verify byte-for-byte integrity before feeding records into agent models or training partitions:
DATA FOUNDRY PROCUREMENT
Data lakes are delivered via secure Cloudflare R2 / AWS S3 presigned download URLs, private bucket replication, or streaming REST/DuckDB endpoints. Partitioned by date and entity domain in Zstandard-compressed Apache Parquet.
Because enterprise dataset requirements vary by historical depth and refresh cadence, Data Foundry access is scoped through transparent metered access quotes. We do not impose automatic recurring subscriptions.
Each lake carries explicit licensing disclosures (U.S. Government Open Data 17 U.S.C. § 105, ODC-By, CC-BY-4.0, or CC0). Proprietary venue feeds (such as Hyperliquid HIP-3/HIP-4) are quarantined and excluded from public syndication.
Yes. ELYRAON provides standard MCP tool adapters allowing Claude Code, Codex, and Cursor agents to query schemas, inspect sample partitions, and execute bounded lookups directly over JSON-RPC 2.0.
Our engineering team reviews all scoping requests within 24 hours. Submit your required entities, date ranges, and delivery protocol via the Quote Intake Form.
AVAILABILITY & LICENSING DISCLOSURES